After synthesis, preparation, and rheological analysis of the hydrogel, the degradation and release behavior of the CXB-loaded hydrogel was evaluated. Furthermore, in an in vitro model in the presence of a proinflammatory stimulus (TNF-α) the ability of the CXB-loaded hydrogels to suppress PGE2 levels in a sustained manner was evaluated. Thereafter, biocompatibility upon subcutaneous injection was studied in mice. Safety of intradiscal application of the loaded and unloaded hydrogel was studied in a canine model of spontaneous mild IVD degeneration. After the hydrogel was shown to be biocompatible and safe, a follow-up dose–response in vivo study was performed in order to determine safety and efficacy of the pNIPAAM MgFe-LDH hydrogel for intradiscal controlled delivery of CXB.
Synthesis and preparation of pNIPAAM MgFe-LDH hydrogels
The poly-N-isopropylacrylamide (pNIPAAM) polymer with sulfonate end group was synthesized as reported previously [21] and the modified synthesis is described in detail in Additional file 1. To formulate the hydrogel, the pNIPAAM polymer was added to the LDH suspension in a vial and subsequently placed on a tube roller mixer for 48 h at room temperature and sterilized by gamma radiation (25 kGy, Isotron Nederland BV, Ede, The Netherlands). The final hydrogel contained 16 wt % pNIPAAM, 3.3 wt % MgFe LDH and water.
Rheological analysis of the pNIPAAM MgFe-LDH hydrogels
The viscoelastic properties of the unloaded pNIPAAm MgFe-LDHs were determined by using an Anton Paar MCR301 rheometer (Anton Paar Ltd., St. Albans, UK) with an oscillatory parallel plate geometry (50 mm diameter) with a constant strain of γ = 0.5 % at a frequency of f = 1 Hz. Temperature was increased from 22 °C to 37 °C at a rate of 15 °C/min. This heating rate was chosen based on calculating the minimum rate of heat transfer based on estimating the energy needed to heat up the hydrogel, by taking into account the surface of the hydrogel and the minimum temperature difference between the LCST and body temperature. LCST is the critical temperature above which the hydrogel undergoes a phase transition from a soluble to an insoluble state. This estimation is described in detail in Additional file 1. The gelling was recorded by measuring the complex shear modulus |G*|, which is a common parameter to determine the strength of a viscoelastic material like a hydrogel. The complex shear modulus |G*| is correlated to the storage modulus (G’) and loss modulus (G”). The storage modulus is a measure of the deformation energy stored in the sample during the shear process (elastic behavior), whereas the loss modulus is a measure of the energy dissipated in the sample during the shear process (viscous behavior), and is lost to the sample afterward (viscous behavior). The relation between these parameters is the following:
$$ \left|G*\right|=\sqrt{{\left(G\hbox{'}\right)}^2+{\left(G"\right)}^2} $$
Hydrogel samples were placed on the lower plate, and the upper plate was lowered to a 0.5 mm gap. The configuration of the rheological setup is shown in Fig. 1a. The viscoelastic properties of the loaded hydrogel were not determined. The CXB concentrations were as low as 10−6 M to 10−4 M, i.e., 0.38–38 mg of celecoxib per liter, or 0.0038–3.8 × 10-5 wt %, and were not influencing the rheological properties.
Degradation and release behavior of CXB-loaded pNIPAAM MgFe-LDH hydrogels
In vitro, the controlled release of CXB from hydrogels was measured in phosphate-buffered saline (PBS) (pH 7.4, 44 mM Na2HPO4, 9 mM NaH2PO4, 72 mM NaCl, 0.02 % wt NaN3) and 0.2 % Tween 80® (polyoxyethylenesorbitan monooleate; Sigma-Aldrich Chemie B.V., Zwijndrecht, The Netherlands). Tween 80® was added to the buffer in order to increase the solubility of CXB [22] and thereby simulate the in vivo situation. CXB-loaded pNIPAAM MgFe-LDH suspension was prepared by adding 6 or 10 mg/ml of CXB to the dispersion and stirring with a stirring bar for 2 days. A volume of 1 ml of CXB-loaded pNIPAAM MgFe-LDH suspension was pipetted into a vial and placed for 30 min at 37 °C to ensure gelation of the hydrogel, and afterward covered with 14 ml warm (37 °C) PBS/Tween 80® solution. The release experiment was performed at 37 °C. At day 1, 2, 5, 8, 15, 22, and 31, 12 ml of the buffer solution was removed in order to analyze CXB and Mg concentrations and 12 ml of fresh buffer was added. CXB concentrations were determined in a volume of 100 μl by using ultra-performance liquid chromatography (UPLC) as described in detail recently by Petit et al. [23]. In vitro degradation was determined by measuring the relative cumulative release of Mg into the medium. Mg concentrations were determined by using a Prodigy High Dispersion Inductively Coupled Optical Emission Spectometry (ICP-OES) system (Leeman Labs, St Charles, IL, USA). Standards were prepared by using multi-element (23 elements in diluted nitric acid) standard solution IV (1000 mg/l) (Merck Millipore, Darmstadt, Germany). A volume of 0.5 ml of the solutions from the degradation experiment was diluted in 100 ml aqueous 1N HNO3, and subsequently diluted again tenfold in 1N HNO3. The effects of CXB loading (10 mg/ml versus 6 mg/ml), LDH content (single versus double) and type of LDH (Mg3Fe versus Mg2.5Fe) of the gels on release behavior and in vitro degradation were also investigated, as well as CXB solubility effects by using a buffer containing PBS and 0.2 % or 2 % Tween 80®.
Controlled release of CXB in vitro
The balance between anabolic and catabolic pathways in articular chondrocytes as well as NP cells can be directed toward catabolism by TNF-α [24, 25]. Bovine articular chondrocytes were used in the in vitro experiments, as they were more easily available in our laboratory. Articular chondrocytes were isolated from bovine carpometacarpal joints by enzymatic digestion overnight with 2 mg/ml collagenase A (Roche Diagnostics Deutschland GmbH, Mannheim, Germany) at 37 °C. Chondrocytes were seeded at 5 × 106 cells/ml density (P0) into cylindrical (diameter and height 6 mm) 2 % agarose (Type VII, Sigma-Aldrich Chemie B.V.) constructs and left to gel at room temperature. The constructs were then cultured in 12-well plates in high-glucose Dulbecco’s modified Eagle’s medium (hgDMEM; Gibco, Life Technologies Europe, Bleiswijk, The Netherlands) with 20 % fetal bovine serum (Greiner Bio-One, Alphen aan Den Rijn, The Netherlands), 0.1 % amphotericin (Sigma-Aldrich Chemie B.V.), 1 % Pen-Strep (Biochrom GmbH, Berlin, Germany), 1 % nonessential amino acids (Lonza, Basel, Switzerland) 1 % essential amino acids (Lonza), and 50 mg/ml ascorbate 2-phosphate (Biochrom). During the first 5 days of culturing, these constructs were stimulated with 10 ng/ml TNF-α to induce an inflammatory response reflected by elevated PGE2 levels (Fig. 2a). A concentration of 1 μM CXB has been described to effectively lower PGE2 levels in osteoarthritic chondrocytes and corresponds with mean pharmacological plasma levels [26]. CXB was dispersed in the pNIPAAM MgFe-LDH mixture at a concentration of 0.1 mg/ml, aiming to establish a concentration of approximately 1 μM CXB per culture medium renewal over the 28-day culture period. Controlled release of the CXB is achieved by dissolution of the CXB crystals present a depot within the hydrogel and diffusion of the solubilized CXB. A volume of 100 μl of the hydrogel suspension was pipetted at the bottom of a 12-well plate and placed in an incubator at 37 °C to ensure gelation of the hydrogels. Subsequently, defined as day 0 of the experiment, cell constructs and culture medium were added to the 12-well plate. For the “bolus injection” of CXB, only cell constructs were placed on the bottom of the well, and CXB was added to the medium every 2 days, starting at day 0, at a concentration of 1 μM. Media were renewed every 2–3 days, collected on days 0, 2, 7, 9, 11, 14, 21, and 28 and stored at −80 °C for analysis of CXB content. For the in vivo experiments based on formulation of higher doses of CXB, in vitro experiments were also carried out with a higher dosage of CXB to evaluate release profiles at higher dosing. To this end, 1 mg of CXB-loaded per ml of hydrogel, aiming to establish a concentration of 10 μM CXB per culture medium renewal was used. Conditioned media were analyzed for CXB content and PGE2 levels. Inhibition of COX-2 activity was determined by measuring PGE2 in culture medium. A colorimetric competitive enzyme immunoassay kit (PGE2 EIA kit, Enzo Life Sciences BVBA, Antwerp, Belgium) was used to determine PGE2 levels in culture medium according to the manufacturer’s instructions.
In vivo biocompatibility in mice after subcutaneous implantation
All animal procedures were approved and performed in accordance with the guidelines set by the Animal Experiments Committee (DEC) of Utrecht University (experiment numbers: DEC 2010.III.03.046; DEC 2012.III.05.046; and DEC 2013.III.02.017). Six healthy female adult (8–10 weeks old) BALB/c mice (Harlan-Olac Ltd., Bicester, UK) were used for testing biocompatibility and biosafety of the pNIPAAM MgFe-LDH polymer hydrogel and seven other biomaterials. Four different biomaterials were injected at least 1 cm apart from each other into the dorsal subcutaneous tissue of each mouse in a randomized fashion. Buprenorphine 100 μg/kg was given intraperitoneally (i.p.) as premedication and analgesic and subsequently all animals were anesthetized with isoflurane via an induction mask (vaporizer setting 2.5 %) in a 1:1 oxygen:air mixture. A blood sample was drawn to perform a white blood cell count and differentiation at day 0, to rule out systemic inflammation. A volume of 200 μl of each biomaterial was injected subcutaneously with a 27G needle under sterile conditions. PBS (200 μl) served as a control. All injection sites were marked with a waterproof marker. Immediately after injection, Dermabond® (Ethicon, Cornelia, GA, USA) was applied to the injection site to prevent leakage and the injection site was heated by an infrared lamp for 1 min. Mice were monitored daily for signs of distress or pain (e.g., lethargy, weight loss, automutilation, and abnormal posture) and injection sites were monitored for inflammation (e.g., swelling, redness, pain, and heat). Three animals were sacrificed 7 days after injection, and three after 28 days. At the end of the experimental period, animals were anesthetized with isoflurane, blood was collected by cardiac puncture for white blood cell count and differentiation, and euthanasia was performed by cervical dislocation. The injection sites were removed for histological analysis. Tissues were fixated in a 4 % neutral-buffered formaldehyde solution (Klinipath B.V., Duiven, The Netherlands) and after fixation routinely embedded in paraffin. Sections of 4 μm were stained with hematoxylin and eosin. Infiltration of inflammatory cells, giant cells, necrosis, neovascularization, fatty infiltration, and the encapsulation of the biomaterial by a fibrotic capsule were histologically assessed as parameters for a biological response at the application site, at 7 and 28 days by a blinded board-certified veterinary pathologist (GG) and the principal investigator (NW) using an Olympus BX41 microscope (Olympus Europa GmbH, Hamburg, Germany).
Intradiscal application of CXB-loaded pNIPAAM MgFe-LDH hydrogels in laboratory beagle dogs
Data from two in vivo studies in beagle dogs were combined and analyzed. Both studies were set up as randomized block designs. In the first study CXB-loaded (7.7 μM) and unloaded pNIPAAM MgFe-LDH hydrogels, a bolus injection of CXB (7.7 μM) and 0.9 % NaCl were intradiscally injected. In other levels two other materials irrelevant to this study were injected. The second study served as a dose–response study, including a 10- and 100-fold higher dosage of CXB (77 μM and 770 μM) in addition to the 7.7 μM dose. For preparation of the CXB-loaded hydrogels, CXB was prepared from a CXB stock solution in ethanol (60 μg/ml) by sterile filtration. Water was added to this ethanolic solution of CXB to obtain a dispersion with small CXB crystals (diameter approximately 1 μm). This dispersion was freeze-dried overnight and the pNIPAAM MgFe-LDH mixture was added and incubated overnight on a tube roller mixer at room temperature.
In total 18 intact female beagle dogs (Harlan, Gannat, France) with a median age of 1.7 years (range 1.3–1.8 years) and a median weight of 8.4 kg (range 6.2–13.8 kg) were used. Nine dogs with a median age of 1.6 years (range 1.3–1.8 years) and a median weight of 8.2 kg (range 6.2–11 kg) were used in the first study. Nine dogs with a median age of 1.7 years (range 1.6–1.8 years) and a median weight of 9.3 kg (range 8.3–13.8 kg) were used in the second study. All dogs underwent general, orthopedic, and neurologic examination by a board-certified veterinary surgeon (BM).
Surgical procedure
To determine the grade of degeneration of the IVDs prior to surgery, magnetic resonance (MR) images of the lumbar vertebral column were obtained in fully anesthetized dogs. A blood sample was drawn from the jugular vein to assess white blood cell count and differentiation, to exclude systemic inflammation. Dogs were placed in a dorsal recumbent position and throughout the complete scan protocol heart rate, respiration rate, temperature, carbon dioxide, and oxygen levels were monitored. The MR imaging was performed using a 0.2 Tesla open magnet (Magnetom Open Viva, Siemens AG, Munich, Germany). All lumbar IVDs were assessed according to the Pfirrmann score by a veterinary radiologist on sagittal T2-weighted fast spin echo (FSE) images (3.0 mm slices, repetition time (TR) 4455 ms, echo time (TE) 117 ms) [27]. Only lumbar IVDs with a Pfirrmann score II were included for injection.
The anesthesia protocol during surgery was similar to the one used during MR scanning. Analgesia was provided by a combination of fentanyl (loading dose 10 μg/kg, 15–20 μg/kg/h continuous rate infusion, c.r.i.) and ketamine (0.5 mg/kg loading dose, 10 μg/kg/min c.r.i.) intravenous (i.v.). Throughout the complete procedure heart rate, respiration, temperature, carbon dioxide, oxygen levels, and blood pressure (noninvasive) were monitored. Surgical sites were prepared according to standard protocol. A detailed description of the surgical procedure has been described previously [28]. Briefly, dogs were positioned in a right recumbent position to expose and inject the T13-L1 until L6-L7 via a left lateral approach. To diminish injury of the iliopsoas muscle and sciatic nerve traction injury, the surgical approach in the second study was adjusted and L6-L7 and L7-S1 were injected via a dorsal approach, while the dogs were positioned in ventral recumbency. In the first study a 100 μl syringe (7638–01 Model 710 RN, Hamilton Company USA, Reno, NV, USA), and in the second study a 100 μl gastight syringe (7656–01 Model 1710 RN) with a 29G needle (25 mm, 12° beveled point; Hamilton Company USA) was used to inject 30 μl of the earlier mentioned compounds through the AF into the NP. The smallest possible needle diameter was chosen to minimize injury to the treated IVDs. Wound closure was performed according to standard protocol. Postoperative pain management in all dogs consisted of methadone 0.3 mg/kg intramuscular (i.m.) quaque (q).6.h. during the first 24 h postoperatively and buprenorphine 20 μg/kg i.m. q.4.h. and/or tramadol 2–5 mg/kg per os (p.o.) q.6.h. the following 7 days. All dogs were treated postoperatively with antibiotics (amoxicillin/clavulanic acid 12.5 mg/kg q.12.h p.o.) during 5 days. Dogs were monitored daily throughout the study by a veterinarian to assess pain symptoms according to the short form of the Glasgow composite pain scale. Dogs that showed signs of pain, received tramadol and/or buprenorphine and/or gabapentin (5 mg/kg p.o. q.12.h). Furthermore, animals were monitored daily by a veterinarian for clinical signs of illness, neurologic deficits and lameness.
Injected substances
In the first study spontaneously degenerated IVDs (Pfirmann grade 2) of the dogs were injected with a volume of 30 μl of NaCl 0.9 % (sham), a bolus of CXB (7.7 μM), a CXB-loaded (7.7 μM) and an unloaded pNIPAAM MgFe-LDH hydrogel. Based on studies in cadaveric spines (unpublished data, N. Willems, and B.P. Meij) a volume of 30 μl could be injected into the NP without substantial resistance. The volume of 30 μl contained 7.7 μM × 10−6 M CXB, to achieve a final concentration of 1 μM (=7.7 × 10−6 × (30 μl gel/230 μl NP volume plus gel)) for the bolus of CXB, in the canine NP of beagle laboratory dogs with a mean weight of 8–9 kg and taking into account the volume of the nucleus (200 μl) [29]. All substances were injected into the IVDs in the T12-L6 spinal segment in a randomized fashion, except for the sham treatment (NaCl 0.9 %), which was injected into T12-T13. An interim statistical analysis was performed after the first study to evaluate treatments and study design. Results were used to perform a new power analysis and to adapt the study design of the second study. In the second study, all substances were administered in a random order within each animal and IVDs of the T12-S1 spinal segment were injected with NaCl 0.9 %, a bolus of CXB (7.7 μM), CXB-loaded (7.7 μM, 77 μM and 770 μM) hydrogels, and an unloaded pNIPAAM MgFe-LDH hydrogel. IVDs adjacent to those injected with hydrogel loaded with the highest dose of CXB (770 μM) remained untreated.
Postmortem collection of materials
Dogs were euthanized 4 weeks postinjection. First, they were sedated with dexmedetomidine 0.04 mg/kg i.v., followed by pentobarbital 200 mg/kg i.v. Immediately after euthanasia, the vertebral column (T12-S1) was harvested by using an electric multipurpose saw (Bosch, Stuttgart, Germany). All muscles were removed and the vertebrae were transected transversely with a band saw (EXAKT tape saw, EXAKT Advanced Technologies GmbH, Norderstedt, Germany), resulting in nine spinal units (endplate–IVD–endplate). These units were then transected sagittally by using a diamond band pathology saw (EXAKT 312 saw; EXAKT diamond cutting band 0.1 mm D64; EXAKT Advanced Technologies GmbH), generating two identical parts. One half was resected with a surgical knife by removing the endplate and the vertebra attached to it on one side, and the remaining IVD tissue was snap frozen in liquid nitrogen stored at −80 °C for biochemical and biomolecular analyses. The other half was photographed (Olympus VR-340, Olympus Europa GmbH) for macroscopic evaluation of the IVD (Thompson score, see below) and stored for 14 days in 50 ml of 4 % buffered formaldehyde at 4 °C for histological analyses.
Histology, COX-2 immunohistochemistry, and TUNEL assay
Samples were decalcified in 35 % formic acid and 6.8 % sodium formate in a microwave oven (Milestone Microwave Laboratory Systems, Bergamo, Italy) overnight at 37 °C, for 7 nights [30] and embedded in paraffin. Five-μm-thick sections were stained with hematoxylin and eosin and with picrosirius red/alcian blue and evaluated according to a grading scheme according to Bergknut et al. [31]. Histological slides were scored blinded and in random order by two independent investigators (NW, AT) using an Olympus BX41 microscope (Olympus Europa GmbH). In case of doubt, samples were also scored by a board-certified veterinary pathologist (GG). All photographs of the macroscopy of the IVD segments were evaluated by two independent blinded investigators (NW, AT) according to the Thompson grading scheme, which has been validated in dogs [32].
Immunohistochemistry for COX-2 was performed on 5 μm sections mounted on KP-plus glass slides. After deparaffinization and rehydration, sections were treated with Dual Endogenous Enzyme Block (Dako S2003, Dako, Carpinteria, CA, USA) for 10 min at room temperature to block nonspecific endogenous peroxidase, followed by two washing steps each of 5 min with Tris-buffered saline with 1 % Tween 20® (TBST). Sections were treated with Tris-buffered saline (TBS) bovine serum albumin (BSA) 5 % solution to block nonspecific binding for 60 min at room temperature, were carefully rinsed and subsequently incubated with a primary mouse anti-human monoclonal COX-2 antibody (Cayman Chemical, Ann Arbor, MI, USA) diluted 1:50 in TBS-BSA 5 % overnight at 4 °C. The following day sections were incubated with peroxidase-labeled polymer (Envision™ anti-mouse K4001, Dako). Antibody binding was visualized by using diaminobenzidine (DAB; Dako). Sections were counterstained with hematoxylin solution (Hematoxylin QS, Vector Laboratories Ltd., Peterborough, UK) and mounted in permanent mounting medium.
A commercial available terminal deoxynucleotidyl transferase dUTP nick-end labeling (TUNEL; Merck Millipore, Darmstadt, Germany) assay was used according to the manufacturer’s instructions to determine apoptosis. The percentage of COX-2-positive and TUNEL-positive chondrocytes over the total number of cells was determined by manual counting in the NP, and in the ventral (VAF) and dorsal AF (DAF), by two blinded independent investigators (NW, SP).
Biomolecular and biochemical analyses
Cryosections (60 μm) of the spinal units were cut with a cryostat (Leica CM1800 cryostat, Leica Microsystems Inc., Bannockburn, IL, USA) and collected on RNAse-free glass slides. The NP and AF tissues were separated and half of the slides were collected in respectively 400 μl and 750 μl Ambion® KDalert™ lysis buffer solution (Life Technologies) in the first study, and in Complete Lysis-M EDTA-free buffer (Roche Diagnostics Nederland B.V., Almere, The Netherlands) in the second study and stored at −80 °C until biochemical analyses were performed. The other half was collected in 300 μl RLT buffer containing 1 % β-mercaptoethanol (Qiagen, Venlo, The Netherlands) and stored at −80 °C until biomolecular analyses were performed.
Quantitative PCR (qPCR) was performed to assess the effects of (controlled release of) CXB at gene expression levels of the NP with regards to: 1) ECM anabolism: aggrecan (ACAN), collagen type II (COL2A1), collagen type I (COL1A1); 2) ECM catabolism: a disintegrin and metalloproteinase with thrombospondin motifs 5 (ADAMTS5), matrix metalloproteinase 13 (MMP13), tissue inhibitor of metalloproteinase 1 (TIMP1); 3) inflammation: tumor necrosis factor alpha (TNFA), interleukin-1β (IL1B), interleukin-6 (IL6) and interleukin-10 (IL10); 4) COX pathway and PGE2 synthesis: prostaglandin E synthase 1 (PTGES1), prostaglandin E synthase 2 (PTGES2), cyclooxygenase 1 (COX1), and cyclooxygenase 2 (COX2); 5) notochordal markers: brachyury (T), cytokeratin-8 (CK8), cytokeratin-18 (CK18); 6) the indirect effect of CXB on Wnt signaling pathway: axin-2 (AXIN2), c-Myc (c-Myc) and cyclin-D1 (CCND1) and 7) apoptosis: caveolin-1 (CAV1), caspase 3 (CASP3), fas ligand (FasL) and Bcl-2 (BCL2). The primer pairs used for qPCR are given in Additional file 2.
The RNeasy Fibrous Tissue Mini Kit (Qiagen, Venlo, The Netherlands) was used to isolate total RNA. To maximize RNA yield, the incubation period with proteinase K was reduced to 5 min. After on-column DNase-I digestion (Qiagen RNase-free DNase kit) RNA was quantified by using a NanoDrop 1000 spectrophotometer (Isogen Life Science B.V., Ijsselstein, The Netherlands). cDNA was synthesized from 20 ng total RNA in a total volume of 15 μl using the iScript™ cDNA Synthesis Kit (Bio-Rad Laboratories B.V., Veenendaal, The Netherlands). qPCR was performed in duplicate using an iCycler CFX384 Touch™ thermal cycler, and IQ SYBRGreen Super mix (Bio-Rad Laboratories). All dog-specific primers were designed in-house using Perlprimer [33] except for MMP13 [34]. Primer specificity was evaluated with BLAST, and the designed amplicon was tested for secondary structures using MFold [35]. Primers were purchased from Eurogentec, Maastricht, The Netherlands. Amplification efficiencies ranged from 80 to 115 %. Relative expression levels were determined by normalizing the cycle threshold (Ct) value of each target gene by the mean Ct value of three reference genes, i.e., glyceraldehyde 3-phosphate dehydrogenase (GAPDH), ribosomal protein S19 (RPS19), and TATA-binding protein (TBP).
To measure glycosaminoglycan (GAG) and DNA content in the NP and AF, samples in Ambion® KDalert™ lysis buffer solution were homogenized in a tube rotator O/N at 4 °C, whereas samples in Complete Lysis-M EDTA-free buffer were homogenized in a TissueLyser II (Qiagen) for 2 × 30 s at 20 Hz. The supernatant and pellet of each NP and AF were digested overnight in a papain buffer (250 μg/ml papain (Sigma-Aldrich) in 50 mM EDTA and 5 mM L-cysteine) at 60 °C. GAG content was quantified by using a 1,9-dimethylmethylene blue assay [36]. The Quant-iT™ dsDNA Broad-Range assay kit in combination with a Qubit™ fluorometer (Invitrogen, Carlsbad, CA, USA) was used in accordance with the manufacturer’s instructions to determine DNA content in the papain-digested NP and AF supernatant and pellets. DNA content in the supernatants of the NP and AF were negligible and therefore not included in the total content.
PGE2 levels were measured with the same colorimetric competitive enzyme immunoassay kit (PGE2 high sensitivity EIA kit, Enzo Life Sciences BVBA) that was used for the in vitro experiments. Both buffers that were used to lyse tissue were validated and standards were diluted in the same lysis buffer as the samples, which did not show strong interference with the performance of the kit. Total GAG content and PGE2 levels were normalized for DNA content in the pellet and were measured in the NP as well as the AF.
Statistical analyses
Power analyses were performed prior to both in vivo studies by using free software [51], and are described in detail in Additional file 3. PGE2/DNA in the NP was considered to be the main read-out parameter. Biochemical and biomolecular data were analyzed by using the R statistical software, package 2.15.2. A linear mixed-effects model was used to analyze the effect of the injected treatments. Factors incorporated into the model as a fixed effect were ‘treatment’ (NaCl, CXB 7.7 μM, CR, CR + 7.7 μM, CR + 77 μM, CR + 770 μM), ‘tissue’ (NP and AF), and their interaction. Random effects ‘dog’ (dog 1–18) and ‘study’ (study 1 and 2) were incorporated to capture the correlation between multiple measurements within one dog. Residual plots and quantile–quantile (QQ)-plots were used to check for possible violations of normality assumptions. In case of violation, data were logarithmically transformed. The Cox proportional hazards regression model was used to estimate the effect of the injected treatments on gene expression levels. Calculations were performed on Ct values for each target gene and the mean Ct value of three reference genes was incorporated into the model as a covariate. If proportional hazard assumptions were violated, the ratio of the Ct values for each target gene to the mean Ct value of the reference genes was used for analysis. Ct values ≥ 40 were right censored. Regression coefficients were estimated by the maximum likelihood method. Model selection was based on the lowest Akaike information criterion (AIC). Confidence intervals were calculated and stated at the 99 % confidence level to correct for multiple comparisons. Differences between treatments were considered significant if the confidence interval did not include 0, whereas hazard ratios were considered significant if the confidence interval did not include 1.